Uploaded image for project: 'Puppet'
  1. Puppet
  2. PUP-3445

cannot find definition Class on node <some node> (when using a specific environment)



    • Bug
    • Status: Closed
    • Major
    • Resolution: Incomplete
    • PUP 3.2.4
    • None
    • Catalog Application
    • None
    • OS: Centos 6.4
      Puppet master is running with Apache + Passenger
      2 HA puppet masters with SRV records



      We've recently moved to a new HA puppet environment using SRV records and 2 puppet masters to server catalogs (and another separate server to serve CA requests).
      All puppet masters are running with Apache + passenger.
      The 2 masters mount the same Netapp share as their /etc/puppet directory so the configuration applies to both.

      We have a few environments, the mainly used are the main environment and the "deployments" environment we're using to deploy new versions of our code.

      Here's the configuration:

      manifest = /etc/puppet/deployments/site.pp
      modulepath = /etc/puppet/rndmodules:/etc/puppet/submodules:/etc/puppet/opmodules
      filetimeout = 10
      	ca = false
      	modulepath = /etc/puppet/testmodules/:/etc/puppet/rndmodules:/etc/puppet/opmodules:/etc/puppet/submodules
      	ssl_client_header = SSL_CLIENT_S_DN
      	ssl_client_verify_header = SSL_CLIENT_VERIFY
      	filetimeout = 10

      After moving to the new HA solution described above we started getting the following error intermittently when running with the "deployments" environment:

      Error: Could not retrieve catalog from remote server: Error 400 on SERVER: Cannot find definition Class on node <current node>.

      it sometimes even work only after the fifth attempt.

      This never happened before when we were using a single puppet master to server both catalogs and CA (Without SRV records of course)

      I saw a couple of threads describing this error, but it didn't lead me anywhere.
      I've added debug to the rack config.ru but it doesn't give any more details.

      I also tried to change the SRV priorities so that the agent's work with a single master, but there was no change.

      Where could the issue be? is it a bug? Can the Netapp share between the masters cause any of these issues? Is it the SRV records solution causing issues?

      Please assist.


        Issue Links



              Unassigned Unassigned
              chen_l87@yahoo.com Chen Leibovich
              Eric Thompson Eric Thompson
              0 Vote for this issue
              2 Start watching this issue



                Zendesk Support